Stability for core::cmp, core::default, alloc::rc, std::task

[brson]

Summary:

• alloc::rc module stable
• Rc type stable
• Functions relating to weak references experimental
• core::cmp module stable
• PartialEq/Eq/PartialOrd/Ord unstable because trait reform will make them change again
• Equiv experimental because there may be better sol'ns
• lexical_ordering deprecated because it can be done trivially with the Ord trait
• min/max stable
• std::task module stable
• TaskBuilder::stdout/stderr experimental because we aren't certain we want to configure the environment this way
• try_future experimental because Future is experimental
• try unstable because the error type might change
• deschedule/failing unstable

The major thing I did differently than previously-discussed is that I made try experimental: there's been discussion that the error type Box<Any + Send> is not sufficient.

Per https://github.com/rust-lang/meeting-minutes/blob/master/Meeting-API-review-2014-07-16.md.

[brson]

This also adds task::name to replace the deprecated task::with_name and calls it stable.

[huonw]

There's no goal to remove this allocation? (Particularly because the name is often/sometimes a &'static str?)

[alexcrichton]

Inspecting the name is often a debugging utility, not done in performance critical contexts, so I don't think there's really any benefit in avoiding an allocation.

The only part that may be performance sensitive (not measured) would be spawning a task with a specific name and whether to special-case the &'static str case, but the actual inspection of the name seems much rarer.

[huonw]

Maybe a logging library would include the task name in logged messages and thus being as performant as possible might be nice? (But yes, seems rare enough to not matter hugely.)

[huonw]

(Sorry for questioning things that have already been discussed!)